awesome-chemistry-datasets

Contributions are very welcome - please follow the guidelines and the Code of Conduct.

text datasets

• BC5CDR:1500 PubMed articles with 4409 annotated chemicals, 5818 diseases and 3116 chemical-disease interactions (named entity recognition)
• BioCreative V: BC5CDR corpus consists of 1500 PubMed articles with 4409 annotated chemicals, 5818 diseases and 3116 chemical-disease interactions.
• BioRxiv XML - Bulk access to the full text of bioRxiv articles for the purposes of text and data mining (TDM) is available via a dedicated Amazon S3 resource.
• ChemTables: 788 chemical patent tables with labels of their content type. Built for semantic classification of table type. Licensed under CC BY NC 3.0.
• Elsevier Corpus: This is a corpus of 40k (40,001) open access (OA) CC-BY articles from across Elsevier’s journals represent the first cross-discipline research of data at this scale to support NLP and ML research.
• Europe PMC - Bulk download of full text and SI of > 5 million articles.
• IUPAC Gold Book
• LibreText: Open-access chemistry textbook.
• MedRxiv XML - Text and data mining is possible via dedicated Amazon S3 resource.
• NLM literature archive: NLM LitArch (NLM Literature Archive) is a digital archive for books, documents, and articles in the fields of life science, medicine, and healthcare at the National Institutes of Health. Also accessible via NCBI bookshelf. See also the NLMChem, a manually annotated full-text resource on chemicals in the biomedical literature. It contains 150 full-text journal articles selected both to be rich in chemical mentions and for articles where human annotation was expected to be most valuable. However, I saw NLM literature archive already on the list but wasn't sure if it included this dataset
• OpenStax Free textbooks, including Chemistry 2e, which is released under CC-BY 4.0.
• PubChemSTM: 281K chemical structure and text pairs
• PubMed central: free full-text archive
• PubMed: abstracts and outlinks
• PubMedQA: answer research questions with yes/no/maybe using abstracts (1k expert labeled, 61.2k unlabeled and 211.3k artificially generated QA instances).
• S2ORC: The Semantic Scholar Open Research Corpus. 81.1M English-language academic papers spanning many academic disciplines largest publicly-available collection of machine-readable academic text). Released under CC BY-NC 4.0.

structures

• COCONUT: is an open source project for Natural Products (NPs) storage, search and analysis.
• Crystallography Open Database: open-access collection of crystal structures of organic, inorganic, metal-organic compounds and minerals, excluding biopolymers. They also derived SMILES for some compounds.
• Enamine HTS collection: 1 930 980 diverse screening compounds (37 billion molecules in 2D and 4.5 billion in 3D)
• GDB: enumeration of molecules according to simple (feasibility and stability) rules
• GNPS: mass spectrometry database with focus on natural products, contains untargeted (unlabelled) data.
• MoNA: mass spectrometry database of real and predicted spectra for known compounds.
• nCov-Group Data Repository: SMILES, fingerprints, descriptors, and images of millions of compounds.
• nmrshiftdb2: is database for organic structures and their nuclear magnetic resonance (NMR) spectra.
• zinc20: ZINC20 library prepared for Deep Docking-accelerated virtual screening
• zinc22: commercially-available compounds for virtual screening

molecular activity prediction benchmark datsets

• MPCD: a benchmark for molecular activty prediction, including both 9 Low-sample size and narrow-scaffold inhibitors datasets(LSSNS) and 30 Higher-sample size and mixed-scaffold inhibitor datasets(HSSMS), each dataset is visulised by TMAP
• MoleculeACE: a benchmark (30 HSSMS datasets in MPCD) for evaluating the predictive performance on activity cliff compounds of machine learning models.

ml structure-property benchmark datasets

• ACNet: a benchmark for Activity Cliff Prediction, 400K Matched Molecular Pairs (MMPs) against 190 targets, including over 20K MMP-cliffs and 380K non-AC MMPs from ChEMBL (version 28).
• Aquasoldb: Curation of nine open source datasets on aqueous solubility. The authors also assigned reliability groups.
• BigSolDB 2.0: Molecular solubility in organic solvents and water in a wide range of temperatures. It contains 103944 experimentally measured solubility values of 1448 organic compounds in 213 solvents reported in the 1595 literature peer-reviewed articles. Initially BigSolDB, posted in this preprint. Recently (2025), updated to BigSolDB 2.0, published in Scientific Data.
• BindingDB: molecular recognition database, contains 2.6M data for 1.1M Compounds and 8.10K Targets (Feb 2023)
• ChEBI-20: 33,010 molecule-description pairs (for molecule captioning task)
• ESol: Water solubility data(log solubility in mols per litre) for common organic small molecules.
• Flashpoint: Sun et al. collected a dataset of the flashpoints of 10575 molecules from academic papers, the Gelest chemical catalogue, the DIPPR database, Lange's Handbook of Chemistry, the Hazardous Chemicals Handbook, and the PubChem database.
• FreeSolv: Experimental and Calculated Small Molecule Hydration Free Energies
• Harvard OPV: "experimental photovoltaic data from the literature, and corresponding quantum-chemical calculations performed over a range of geometries, each with quantum chemical results using a variety of density functionals and basis sets"

Target identification data

• Open Targets: is a large-scale resource that uses human genetics and genomics data for systematic drug target identification and prioritization.
• Probes & Drugs Portal: is an interactive, open data resource for chemical biology. Overview of libraries of bioactive compounds (e.g., ChEMBL, Guide to PHARMACOLOGY), including commercial screening libraries.

Pharmacology & ADME & Metabolism

• SIDER dataset: The SIDER Side Effect Resource represents an effort to aggregate dispersed public information on side effects. As there is no such resource exist in machine-readable form despite the importance of research on drugs and their effects. Creation of this resource was is related to paper (Campillos, Kuhn et al., Science,